cs448f: image processing for photography and vision

Click here to load reader

Post on 25-Feb-2016

48 views

Category:

Documents

6 download

Embed Size (px)

DESCRIPTION

CS448f: Image Processing For Photography and Vision. Blending and Pyramids. Blending. We’ve aligned our images. What now? Averaging Weighted averaging min/max/median. Noise reduction by Averaging. 2 Shots. 4 Shots. 8 Shots. 16 Shots. Noise Reduction by Averaging. - PowerPoint PPT Presentation

TRANSCRIPT

Fast Filtering & Joint Image Filtering

CS448f: Image Processing For Photography and VisionBlending and PyramidsBlendingWeve aligned our images. What now?AveragingWeighted averagingmin/max/medianNoise reduction by Averaging

2 Shots

4 Shots

8 Shots

16 Shots

Noise Reduction by AveragingWere averaging random variables X and YBoth have variance S2Variance of X+Y = 2S2Std.Dev. of X+Y = sqrt(2) . SStd.Dev of (X+Y)/2 = sqrt(2)/2 . SIe, every time we take twice as many photos, we reduce noise by sqrt(2)

Noise Reduction by AveragingAverage 4 photos: noise gets reduced 2xAverage 8 photos: noise gets reduced 3xAverage 16 photos: noise gets reduced 4x

Noise Reduction by Median(demo)

Median v Average

Median v AverageCan we identify the bad pixels?Theyre unlike their neighboursInstead of averaging, weighted average where weight = similarity to neighboursWeighted Average

Can we identify the bad pixels?Theyre unlike their neighboursInstead of averaging, weighted average where weight = similarity to neighboursFavors blurriness Input

Other uses of MedianRemoving Transient Occluders(live demo)(Gates demo)(surf demo)Panorama Stitching

Panorama Stitching

Panorama Stitching

Panorama Stitching

Panorama Stitching

Multiple Exposure Fusion

Multiple Exposure Fusion

Multiple Exposure Fusion

Multiple Exposure Fusion

Multiple Exposure Fusion

Multiple Exposure Fusion

Multiple Exposure Fusion

Multiple Exposure Fusion

Multiple Exposure Fusion

Multiple Exposure Fusion

Multiple Exposure Fusion

Focus Fusion

Focus Fusion

Focus Fusion

Focus Fusion

Focus Fusion

PyramidsWeve been breaking images into two terms for a variety of appsCoarse + Fine

More generally we can break it into many terms:Very coarse + finer + finer ... + finest.PyramidsWe can do this by blurring more and more:

PyramidsAnd then (optionally) taking differences

--PyramidsThe coarse layers can be stored at low res.

GaussianPyramidLaplacianPyramidPyramidsHow much memory does this use?

Pyramid Uses:Sampling arbitrarily sized Gaussians

Equalizing an imageThe different levels represent different frequency rangesWe can scale each frequency level and recombine

Blending multiple imagesPyramid BlendingKey Insight:Coarse structure should blend very slowly between images (lots of feathering), while fine details should transition more quickly.

More robust to tricky cases than plain old compositingInputs:

Compositing: Hard Mask

Compositing: Soft Mask

Multi-Band Blending

Exposure Fusionhttp://research.edm.uhasselt.be/~tmertens/papers/exposure_fusion_reduced.pdf